At times, progress occurs so quickly that it’s difficult to separate science fiction from real life. Just five decades ago, computers were massive, unwieldy machines running on punch cards and primitive circuits. Today, a single smartphone has more processing power than the computer used on the Apollo missions.
AI has benefited greatly from this explosion in computing power and capability. Today, highly complex deep learning algorithms, patterned on the structure of the human brain, can master Go, trade stocks, and even write Harry Potter novels (though admittedly not very good ones). Given this versatility, some fear that deep learning AIs will reshape our economy by force, rendering hundreds, if not thousands, of occupations obsolete. It seems the world will no longer need humans. But is this really the case? No two occupations are alike, and the Great AI Reckoning will not affect all industries (and certainly not all employees) equally.
Before we go any further, it’s important to figure out the automatability of the medical profession. How likely is it that an algorithm could replace a doctor? At some point in the future, will hospitals force us to enter our symptoms into touchscreens and wait for a disembodied electronic voice to give us a diagnosis?
Thankfully, that doesn’t seem likely. Based on research carried out by Oxford University and NPR, physicians and surgeons only have a 0.4 percent chance of falling victim to automation. Overall, most of the medical professions seem to have a far lower chance of automation than others. Physician assistants, for instance, have an automatability rating of 14.5 percent, while tax preparers have an automatability rating of 98.7 percent.
Any automatability study will take a close look at the duties associated with a job. At its core, what sort of tasks will a worker perform? Will they spend their time on routine tasks that a machine can easily break down into steps and replicate, or are they required to negotiate, exercise empathy, and use creativity and lateral thinking?
Doctors, needless to say, are the latter. Diagnosing diseases, performing surgery, and prescribing medicine aren’t simply complex, life-threatening tasks — they also require a good deal of empathy. Thus far, computers fail at this crucial requirement (hence Silicon Valley’s fear of AI). Though efforts are underway to teach computers empathy, it’s unclear whether it will ever be possible to build a computer that can understand the depths of human emotion.
When it comes to AI’s impact on medicine, I think we’ll find the result to be both significant and subtle.
1. AIs will catch mistakes
Even though computers will never replace flesh-and-blood physicians, artificial intelligence still has a place in medicine as a partner.
Think about what is required of a doctor: an understanding of biochemistry, such as new drugs, existing ones, and how such substances interact with individual patients; general information about each patient’s medical history, including whether they have any pre-existing conditions or risk factors that could be exacerbated by surgery or treatment; and a deep knowledge of diseases and conditions, which often evolve incredibly quickly.
The fact of the matter is that these requirements aren’t just conflicting ones, they’re humanly impossible. After all, a human brain consists of about one billion neurons, and each neuron has around 1,000 connections (for a total of one trillion connections). As impressive as this may sound, it’s not much. One brain has several gigabytes of working memory. Anything else is not so easily recalled.
Fortunately, AIs have no such problem. IBM’s Watson, for instance, can comb through millions of pages of data, read countless medical articles, and far surpass any human doctor in its breadth and scale of knowledge. Even if a doctor may forget that a patient’s unique biology makes them susceptible to a certain drug’s side effects, an AI won’t. And even if an overworked medical resident may miss a clue, an AI won’t.
2. AIs can help with rare conditions
On a related note, the powerful network of an AI will revolutionize the treatment of rare diseases. Yes, individual AIs are powerful, intelligent programs. However, when they’re networked together, they are unstoppable. In this configuration, they can draw from each others’ insights, see where one person went wrong, and devise innovative fixes.
Such tools already exist. A Wired feature discusses Modernizing Medicine, an AI-powered database that helps practitioners tap into knowledge from a database of 3,700 providers and over 14 million patient visits. Based on a technology similar to Amazon’s notoriously powerful recommendation engine, Modernizing Medicine mines data, recommends treatments, and, through the power of the network, helps busy doctors tackle an unfamiliar, threatening disease.
3. AIs will assist with surgery
It’s no wonder the drama depicted on shows like Grey’s Anatomy is so compelling. Surgery is incredibly complex, requires intense (and intensive) specialist training, and is, quite literally, a matter of life and death. AI could help reduce some of this drama.
When paired with augmented reality programs, which overlay digital cues and images atop real ones (think Pokemon Go), AI can provide surgeons with real-time information. This ranges from dividing the brain into various regions for neurosurgery to laying MRI scans and other imagery on top of a patient’s body (giving doctors X-ray vision). Complex software will power advanced hardware, giving doctors an extra safety net and providing some peace of mind for patients.
4. AIs will predict disease
A critical benefit of AI comes from its strength in gathering and analyzing reams of data and drawing conclusions from its analysis. Who is more likely to get cancer? What are the risk factors that make a patient more susceptible to, say, heart attacks as opposed to strokes?
Google, the king of analytics, has already jumped on the bandwagon. Several years ago, Google created its Baseline study, a comprehensive, ambitious undertaking that involved thousands of volunteers and 100 specialists in different medical fields. As the name suggests, the goal of the study was to establish a sort of baseline for human health from which algorithms and researchers could isolate biological clues that could predispose a person to specific illnesses.
Today, the Baseline study has continued under the banner of Verily, a division of Alphabet (Google’s parent company) and is set to expand in both scope and resources. In the near future, it’ll be easy to imagine a time when non-communicable diseases (strokes, cancers, heart attacks) or hereditary conditions are identified from a single visit to the doctor’s office. Not only can patients see their chances of contracting a specific disease, but doctors can also help their patients preempt these conditions with a clear-sighted, long-term plan of action.
It’s important to note that while AI will certainly revolutionize our relationships with medicine, it is far more likely to do so in a subtle, understated way. After all, AI’s most promising changes are related to systems and procedures in the form of back-end interfaces — not in, say, talking screens. Don’t be fooled, however: Even if most of the change occurs off-screen, medical practice will change for the better. Health care will become more accurate, more comprehensive, and cheaper over time, which is welcome news for everyone.